Research Experience

Research Intern

Advisor: Prof. Sehoon Kim

• Researching NVFP4 W4A4 quantization and its performance in looped transformer models.
• Optimizing inference kernels to improve inference speedups.

Research Project

Initiated and led the research; co-authored the paper with Prof. Sehoon Kim (KAIST).

• Built a resource scheduler for on-device LLMs, achieving 2.0× retrieval throughput over a fixed scheduling baseline under the same latency constraint.
• Paper accepted to the NeurIPS 2026 Workshop on On-Device Intelligence.

Research Intern

Advisor: Prof. Falko Kuester

• Evaluated a time-series foundation model for zero-shot anomaly detection on real-world operational data.
• Evaluated proxy anomalies and alert-volume tradeoffs; found that forecast residuals did not reliably predict genuine failures.

Undergraduate Researcher

Advisor: Prof. Sung-Eui Yoon

• Evaluated the semantic robustness of vision-language-action (VLA) policies in simulation under 11 meaning-preserving perturbations.
• Average success fell from 60% to 25%, with structural edits causing the largest drop.

Research Intern

• Built a real-time brain–computer interface mapping EEG signals to robotic-arm commands using PyTorch.